Doctors, AI makers and hospitals share liability risks in new
The American Medical Association has established clear roles for physicians, AI developers, and healthcare organizations to prevent patient harm from clinical AI tools, as outlined in a new issue brief. The guidelines seek to avoid disputes over liability while enhancing patient safety as AI integration in medicine expands.
Physicians Must Exercise Judgment Without Unfair Liability
Doctors retain responsibility for applying clinical judgment when using AI tools, including evaluating AI-generated suggestions. However, the AMA argues physicians should not bear liability for errors outside their control—such as when an AI system ignores critical patient data or when physicians lack insight into decision-making processes.
An AMA survey from earlier this year revealed that one-third of physicians cited the need for clear liability frameworks as the primary reason for adopting AI in healthcare. The new guidelines clarify accountability for each stakeholder, reducing the risk of physicians facing blame for AI failures.
Developers Must Ensure Safety and Transparency
The AMA’s framework places primary responsibility on AI developers to create tools that provide accurate, safe recommendations. Unlike current practices relying on vague disclaimers, the association insists developers must disclose detailed information about a tool’s intended use, performance limitations, and potential risks. Without this transparency, physicians cannot properly assess the reliability of AI outputs.
Jennifer Goldsack, head of the AMA Center on Digital Health and AI, emphasized that disclaimers alone are inadequate. The guidelines instead require identifying specific risks, assigning accountability to the party best positioned to prevent harm, and ensuring users have the necessary information to address those risks.
Related Post: Lawmaker proposes overhaul to fix broken medical arbitration system
Healthcare Organizations Must Vet, Train, and Monitor AI Systems
Healthcare organizations deploying AI tools, including hospitals, health systems, and physician practices, must select appropriate tools, provide training, and monitor performance over time.
Organizations must also establish processes to detect and respond to AI errors, particularly as model accuracy degrades. The framework stresses that without proper oversight, health systems risk exposing patients to harm by relying on untested or poorly maintained AI tools.
The AMA’s push for clearer accountability also addresses patient recourse. If AI causes harm, patients should have defined pathways to seek compensation from the responsible party, whether developer, deployer, or physician, though prevention remains the primary focus.
Growing AI Autonomy Raises Liability Concerns
As clinical AI becomes more autonomous, disputes between developers and providers over responsibility for errors are intensifying. In recent comments to the Food and Drug Administration, the AMA and other provider groups argued that AI developers, not healthcare providers, should bear the burden of monitoring generative AI-enabled medical devices. They cautioned that placing this responsibility on providers, especially smaller or rural practices, would create unsustainable legal and operational challenges.
The FDA’s proposed regulations for generative AI in healthcare reflect this debate, with providers advocating for liability limits while developers face increasing pressure to meet safety standards. The AMA’s framework aims to align these expectations, ensuring each party fulfills its role in reducing risks before they affect patients.
